
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
import plotly.graph_objects as go
import matplotlib.gridspec as gridspec
import scikit_posthocs as sp
import statsmodels.api as sm

from ipywidgets import Dropdown, IntSlider, interact
from scipy import stats
from scipy.stats  import t 
from plotly.subplots import make_subplots

from category_encoders import OneHotEncoder

from sklearn.utils import resample
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.metrics import confusion_matrix,ConfusionMatrixDisplay, accuracy_score, classification_report, roc_auc_score, precision_score , recall_score

from sklearn.preprocessing import LabelEncoder, OneHotEncoder
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline, make_pipeline

(1470, 35)
0

Text(0.5, 0.98, 'Overall atrrition')

(86.19025367670434, 15.50731305586545, 2.752481638050657e-15)

(72.52901310667391, 9.487729036781154, 6.634684715458909e-15)

(-6.203935765608938, 1.9615812836543436, 7.14736398535381e-10)

(3.201782929420171, 0.04097409724987449, 3.001858137213299)

array(['Sales', 'Research & Development', 'Human Resources'], dtype=object)

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)


import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
import plotly.graph_objects as go
import matplotlib.gridspec as gridspec
import scikit_posthocs as sp
import statsmodels.api as sm

from ipywidgets import Dropdown, IntSlider, interact
from scipy import stats
from scipy.stats  import t 
from plotly.subplots import make_subplots

from category_encoders import OneHotEncoder

from sklearn.utils import resample
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.metrics import confusion_matrix,ConfusionMatrixDisplay, accuracy_score, classification_report, roc_auc_score, precision_score , recall_score

from sklearn.preprocessing import LabelEncoder, OneHotEncoder
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline, make_pipeline

(1470, 35)
0

Text(0.5, 0.98, 'Overall atrrition')

(86.19025367670434, 15.50731305586545, 2.752481638050657e-15)

(72.52901310667391, 9.487729036781154, 6.634684715458909e-15)

(-6.203935765608938, 1.9615812836543436, 7.14736398535381e-10)

(3.201782929420171, 0.04097409724987449, 3.001858137213299)

array(['Sales', 'Research & Development', 'Human Resources'], dtype=object)

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)


import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
import plotly.graph_objects as go
import matplotlib.gridspec as gridspec
import scikit_posthocs as sp
import statsmodels.api as sm

from ipywidgets import Dropdown, IntSlider, interact
from scipy import stats
from scipy.stats  import t 
from plotly.subplots import make_subplots

from category_encoders import OneHotEncoder

from sklearn.utils import resample
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.metrics import confusion_matrix,ConfusionMatrixDisplay, accuracy_score, classification_report, roc_auc_score, precision_score , recall_score

from sklearn.preprocessing import LabelEncoder, OneHotEncoder
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline, make_pipeline


df = pd.read_csv("Employee-Attrition.csv")


df.head()


def ReplaceIt(i,st=""):
    df[st]=df[st].replace(1-i,'Low')
    df[st]=df[st].replace(2-i,'Medium')
    df[st]=df[st].replace(3-i,'High')
    df[st]=df[st].replace(4-i,'Very High')

CatList = ["EnvironmentSatisfaction", "JobSatisfaction","RelationshipSatisfaction","WorkLifeBalance","StockOptionLevel"]

i=0
for st in CatList:
    if st =="StockOptionLevel":
        i = i+1
    ReplaceIt(i,st)


df["JobLevel"] = df["JobLevel"].replace(
    {1:"Entry Level",
     2:"Junior Level",
     3:"Mid Level",
     4:"Senior Level",
     5:"Executive Level"}
)


df["Education"] = df["Education"].replace(
    {1:"Below College",
     2:"College",
     3:"Bachelor",
     4:"Master",
     5:"Doctor"}
)


df['PerformanceRating']=df['PerformanceRating'].replace(3,'Excellent')
df['PerformanceRating']=df['PerformanceRating'].replace(4,'Outstanding')


df.describe()

(1470, 35)
0

Text(0.5, 0.98, 'Overall atrrition')

(86.19025367670434, 15.50731305586545, 2.752481638050657e-15)

(72.52901310667391, 9.487729036781154, 6.634684715458909e-15)

(-6.203935765608938, 1.9615812836543436, 7.14736398535381e-10)

(3.201782929420171, 0.04097409724987449, 3.001858137213299)

array(['Sales', 'Research & Development', 'Human Resources'], dtype=object)

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)


df.describe()


print(df.shape)
print(df.duplicated().sum())

tabel = pd.DataFrame({
    'Unique':df.nunique(),
    'Null':df.isna().sum(),
    'Type':df.dtypes.values
})
display(tabel)

(1470, 35)
0

[
    'EmployeeCount',
    'Over18',
    'StandardHours',
    'EmployeeNumber'
]


df.drop(columns=['EmployeeCount', 'Over18', 'StandardHours', 'EmployeeNumber'], inplace=True)


df.hist(figsize=(15,15))
plt.tight_layout()
plt.show()


fig, axes = plt.subplots(1,2,figsize=(13,6))
sns.countplot(x="Attrition", data=df,ax =axes[0])
for container in axes[0].containers:
    axes[0].bar_label(container)
axes[1].pie(df["Attrition"].value_counts().values, labels=df["Attrition"].value_counts().index,autopct='%1.2f%%',explode=[0, 0.1])
axes[1].legend()
plt.suptitle("Overall atrrition",fontsize=20)

Text(0.5, 0.98, 'Overall atrrition')


# apply t test between two groups seperated by "group" column in given value

def TTest(val="", group=""):
    uni = df[group].unique()
    result = stats.ttest_ind(df[df[group]==uni[0]][val] , df[df[group]==uni[1]][val])
    return result.statistic,stats.t.ppf(1-.05/2,result.df), result.pvalue


# apply t test between all groups seperated by "group" column in all numerical value

def ApplyTTest(group=""):
    T_Score = {}
    CriticalT = {}
    p_values = {}

    for col in df.select_dtypes("number").columns:
        T_Score[col],CriticalT[col],p_values[col] =TTest(col, group)
    columns = list(T_Score.keys())
    values = list(T_Score.values())
    critical = list(CriticalT.values())

    test_df = pd.DataFrame(
        {"Features":columns,
        "T Value":values,
        "Critical Value":critical}
                          )

    test_df["P_value"] =  [format(p, '.20f') for p in list(p_values.values())]
    test_df["P_value"] = test_df["P_value"].astype(float)
    test_df["Result"] = test_df["P_value"].map(lambda x:"Accept" if x > 0.05 else "Reject")
    return test_df, p_values


# calculate max estimation error in num between cat groups 

def CalcMEE(cat="", num="", t=1.96):
    
    #t= 1.96 -> 0.95 CI
    n ={}
    std= {}
    EstError ={}
    Average = {}
    for col in df[cat].unique():
        group = df[df[cat] == col][num]
        Q1, Q3 = group.quantile([0.25, 0.75])
        IQR = Q3 - Q1
        upper = Q3 +1.5*IQR
        lower = Q1 - 1.5*IQR
        mask = group.between(lower, upper)
        group = group[mask]
        n[col] = len(group)
        std[col] = group.std()
        Average[col] = group.mean()
        t= stats.t.ppf(q=1-.05/2,df=len(group)-1)        
        EstError[col] = t* std[col] / np.sqrt(n[col])
    columns = list(n.keys())
    N = list(n.values())
    Std = list(std.values())
    E = list(EstError.values())
    Avg = list(Average.values())
    test_df = pd.DataFrame(
        {cat:columns,
        "Total number":N,
        "Standard Deviation":Std,
         "Mean":Avg,
        "Max Error":E,
        }
                          )
    test_df["Max Error"] = ["+/- "+ str(round(e,4)) for e in E]
    return test_df


# same as CalcMME but divide groups by attrition also 

def SpecialCalcMEE(cat="", num="", t=1.96):
    
    #t= 1.96 -> 0.95 CI
    n ={}
    std= {}
    EstError ={}
    Average = {}
    columns=[]
    
    Att =df[df["Attrition"]=="Yes"]
    NoAtt =df[df["Attrition"]=="No"]

    for col in df[cat].unique():
        group = Att[Att[cat] == col][num]
        colname = col+"-Yes"
        columns.append(colname)
        Q1, Q3 = group.quantile([0.25, 0.75])
        IQR = Q3 - Q1
        upper = Q3 +1.5*IQR
        lower = Q1 - 1.5*IQR
        mask = group.between(lower, upper)
        group = group[mask]
        n[colname] = len(group)
        std[colname] = group.std()
        Average[colname] = group.mean()
        t= stats.t.ppf(q=1-.05/2,df=len(group)-1)
        EstError[colname] = t* std[colname] / np.sqrt(n[colname])
        
        
        group = NoAtt[NoAtt[cat] == col][num]
        colname = col+"-No"
        columns.append(colname)
        Q1, Q3 = group.quantile([0.25, 0.75])
        IQR = Q3 - Q1
        upper = Q3 +1.5*IQR
        lower = Q1 - 1.5*IQR
        mask = group.between(lower, upper)
        group = group[mask]
        n[colname] = len(group)
        std[colname] = group.std()
        Average[colname] = group.mean()
        t= stats.t.ppf(q=1-.05/2,df=len(group)-1)
        EstError[col] = t* std[colname] / np.sqrt(n[colname])
        
    N = list(n.values())
    Std = list(std.values())
    E = list(EstError.values())
    Avg = list(Average.values())
    
    test_df = pd.DataFrame(
        {cat:columns,
        "Total number":N,
        "Standard Deviation":Std,
         "Mean":Avg,
        "Max Error":E,
        }
                          )
    test_df["Max Error"] = ["+/- "+ str(round(e,4)) for e in E]
    return test_df


#Plot Attrition per department

def Plot(dep=""):
    Depatment=df.groupby(['Department','Attrition'])["Age"].count().reset_index(name='Counts')
    Depatment=Depatment[Depatment["Department"]==dep]
    fig, axes = plt.subplots(1,2,figsize=(13,6))
    sns.barplot(x="Attrition",y="Counts", data=Depatment,ax =axes[0])
    for container in axes[0].containers:
        axes[0].bar_label(container)
    axes[1].pie(Depatment["Counts"], labels=Depatment["Attrition"].value_counts().index,autopct='%1.2f%%',explode=[0, 0.1])
    axes[1].legend()
    plt.suptitle(dep,fontsize=20)


#Do Chi square test between two given categories features 

def ChiSquared(cat1="", cat2="Attrition"):
    contingency_table = pd.crosstab(df[cat1], df[cat2])
    chi2, p_value, deg, _ = stats.chi2_contingency(contingency_table)
    CriticalChi = stats.chi2.ppf(0.95, deg)
    return chi2,CriticalChi, p_value


#plot p values of results of ANOVA test or chi 

def plotP(p_values,st="", ISchi = False, tit =""):
    plt.figure(figsize=(12,6))
    keys = list(p_values.keys())
    values = list(p_values.values())
    sorted_value_index = list(reversed(np.argsort(values)))
    sorted_dict = {keys[i]: values[i] for i in sorted_value_index}
    keys = list(sorted_dict.keys())
    values = list(sorted_dict.values())
    
    sns.barplot(x=keys, y=values, palette='rocket')
    if tit =="": 
        if ISchi:
            plt.title(f"P_scores Comparison according Chi Square test between Categorical variables and Atrrition")
        else:
            plt.title(f"Anova Test P_score Comparison within {st} Groups")
    else:
        plt.title(tit)

    plt.axhline(y=0.05,color='red',linestyle ="--", label="P_value = 0.05")
    plt.legend()

    plt.xticks(rotation=90)

    for index,value in enumerate(values):
        plt.text(index,value,round(value,3), ha="center", va="bottom")
    plt.show()


# apply ANOVA test to all numerical values of groups based on give cat feature st

def SingleANOVA(st="", num=""):
    f_scores = {}
    p_values = {}

    Arg = []

    for col in df[st].unique():
        Arg.append(df[num][df[st] == col])

    f_score, p_value = stats.f_oneway(*Arg)
    dof_within = len(df) - len(Arg) 
    dof_between = len(Arg) -1 
    CriticalF = stats.f.ppf(1-0.05,dof_between,dof_within )
    
    return f_score,p_value, CriticalF


# same as SingleANOVA function but it divide groups by attrition and the givin st 

def SpecialSingleANOVA(st="", num=""):
    f_scores = {}
    p_values = {}

    Arg = []
    
    Att = df[df["Attrition"]=="Yes"]
    for col in Att[st].unique():
        Arg.append(Att[num][Att[st] == col])

        
    NoAtt = df[df["Attrition"]=="No"]
    for col in NoAtt[st].unique():
        Arg.append(NoAtt[num][NoAtt[st] == col])
    
    f_score, p_value = stats.f_oneway(*Arg)
    dof_within = len(df) - len(Arg) 
    dof_between = len(Arg) -1 
    CriticalF = stats.f.ppf(1-0.05,dof_between,dof_within )
    
    return f_score,p_value, CriticalF


# apply ANOVA test to all numerical values of groups based on give cat feature st

def ANOVA(st=""):
    f_scores = {}
    p_values = {}
    CriticalF = {}


    for num in df.select_dtypes("number").columns:
        Arg = []
        
        f_scores[num],p_values[num], CriticalF[num] = SingleANOVA(st, num)
    
    plotP(p_values,st)
    return f_scores,p_values


# same as ANOVA function but it divide groups by attrition and the givin st 

def SpecialANOVA(st=""):
    f_scores = {}
    p_values = {}
    CriticalF = {}


    for num in df.select_dtypes("number").columns:
        Arg = []
        
        f_scores[num],p_values[num], CriticalF[num] = SpecialSingleANOVA(st, num)
    
    plotP(p_values,st)
    return f_scores,p_values


# plot error bar between given p values with p = 0.5

def PlotErrors(p_values,st ="" ,ISchi = False):
    plt.figure(figsize=(12,6))

    x= list(p_values.keys())
    y= list(p_values.values())
    y_errormin =len(x) *[0.5]
    y_errormax =len(x) *[0.45]

    y_error =[y_errormin, y_errormax]

    plt.errorbar(x, y,
                 yerr = y_error,
                 fmt='o',capsize=10)
    plt.xticks(rotation=90)

    plt.axhline(y=0.5,color='red')
    if ISchi:
        plt.title(f"Chi Square P_score Error bar between Categorical features and Attrition")
    else:
        plt.title(f"ANOVA Test P_score Comparison within {st} Groups")


# bar plot for chi square results

def PlotChi(keys,values):
    plt.figure(figsize=(12,6))
    sorted_value_index = list(reversed(np.argsort(values)))
    sorted_dict = {keys[i]: values[i] for i in sorted_value_index}
    keys = list(sorted_dict.keys())
    values = list(sorted_dict.values())
    
    sns.barplot(x=keys, y=values, palette="mako")
    plt.xticks(rotation=90)
    plt.title("Chi2 Statistic Value of each Categorical Columns with Attrition",fontweight="black")
    for index,value in enumerate(values):
        plt.text(index,value,round(value,2),ha="center",va="bottom")
    plt.show()


## add percentage column to given Df to use in annot bar plot

def AddPercentage(Df, st=""):
    lenght = len(Df[st])
    Df["percent"]=lenght*[""]
    i=0
    while i <lenght :
        Sum = Df.iloc[i,2] +Df.iloc[i+1,2]
        prc1 = round((Df.iloc[i,2] / Sum) * 100 , 2)
        Df.iloc[i,3] = str(prc1) + ' %'
        prc2= round( (Df.iloc[i+1,2] / Sum )* 100,2)
        Df.iloc[i+1,3] = str(prc2) + ' %'
        i =i+2 
    return Df


#plot four beside pie charts

def PlotPies(st=""):
    bus=df.groupby([st,'Attrition'],as_index=False)['Age'].count()
    bus.rename(columns={'Age':'Count'},inplace=True)
    fig=go.Figure()
    fig = make_subplots(rows=1, cols=4,
                        specs=[[{"type": "pie"}, {"type": "pie"},{"type": "pie"}, {"type": "pie"}]],
                        subplot_titles=('Very High', 'High','Medium','Low'))

    fig.add_trace(go.Pie(values=bus[bus[st]=='Very High']['Count'],labels=bus[bus[st]=='Very High']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=1)
    fig.add_trace(go.Pie(values=bus[bus[st]=='High']['Count'],labels=bus[bus[st]=='High']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=2)
    fig.add_trace(go.Pie(values=bus[bus[st]=='Medium']['Count'],labels=bus[bus[st]=='Medium']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=3)
    fig.add_trace(go.Pie(values=bus[bus[st]=='Low']['Count'],labels=bus[bus[st]=='Low']['Attrition'],pull=[0,0.1],showlegend=True)
                  ,row=1,col=4)

    fig.update_layout(title_x=0.5,template='simple_white',showlegend=True,
                      legend_title_text="Attrition",title_text=f"<b style='color:black; font-size:100%;'>Employee Attrition based on {st}",
                      font_family="Times New Roman",title_font_family="Times New Roman")
    fig.show()


for dep in set(df["Department"]):
    Plot(dep)


 pd.crosstab(df['Department'], df['Attrition'])#.reset_index()


chi2,CriticalChi, p_value = ChiSquared("Department")


f_scores,p_values = ANOVA("Department")


sp.posthoc_ttest(df, val_col='MonthlyIncome', group_col='Department')


jobRole = df.groupby(['JobRole', 'Attrition'])["Age"].count().reset_index(name='Counts')
jobRole = AddPercentage(jobRole,"JobRole")

(86.19025367670434, 15.50731305586545, 2.752481638050657e-15)

(72.52901310667391, 9.487729036781154, 6.634684715458909e-15)

(-6.203935765608938, 1.9615812836543436, 7.14736398535381e-10)

(3.201782929420171, 0.04097409724987449, 3.001858137213299)

array(['Sales', 'Research & Development', 'Human Resources'], dtype=object)

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)

(1.6680187953544354, 1.9615812836543436, 0.0955252620565195)

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)


jobRole = df.groupby(['JobRole', 'Attrition'])["Age"].count().reset_index(name='Counts')
jobRole = AddPercentage(jobRole,"JobRole")


px.bar(jobRole, y ="Counts", x ="JobRole",color="Attrition", text='percent')


ChiSquared("JobRole")

(86.19025367670434, 15.50731305586545, 2.752481638050657e-15)


byJobLevel = df.groupby(['JobLevel', 'Attrition'])["Age"].count().reset_index(name='Counts')
byJobLevel = AddPercentage(byJobLevel,"JobLevel")
order = ["Entry Level","Junior Level","Mid Level","Senior Level","Executive Level"]
px.bar(byJobLevel, y ="Counts", x ="JobLevel",color="Attrition", text='percent',category_orders= {'JobLevel':order})


ChiSquared("JobLevel")

(72.52901310667391, 9.487729036781154, 6.634684715458909e-15)


f_scores, p_values =ANOVA("JobRole")


f_scores, p_values =ANOVA("JobLevel")

[
    "Age"                    , "MonthlyIncome"
    "NumCompaniesWorked"     , "TotalWorkingYears"
    "YearsAtCompany"         , "YearsInCurrentRole"
    "YearsSinceLastPromotion", "YearsWithCurrManager"
]


px.box(df, x="JobLevel", y="Age", title="Age distribution by different job roles")


plt.figure(figsize=(13,6))
plt.subplot(1,2,1)
sns.histplot(x="MonthlyIncome", hue="Attrition", kde=True,data=df )
plt.title("Employee Attrition by Monthly Income")

plt.subplot(1,2,2)
sns.boxplot(x="Attrition",y="MonthlyIncome",data=df)
plt.title("Employee Attrition by Monthly Income")
plt.tight_layout()
plt.show()

(-6.203935765608938, 1.9615812836543436, 7.14736398535381e-10)

(3.201782929420171, 0.04097409724987449, 3.001858137213299)

array(['Sales', 'Research & Development', 'Human Resources'], dtype=object)

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)

(1.6680187953544354, 1.9615812836543436, 0.0955252620565195)

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)

(1.3921381802920636, 1.9615812836543436, 0.16409141231818586)

(14.58158957406234, 2.3213794446188195e-09, 2.6109723453486713)


plt.figure(figsize=(13,6))
plt.subplot(1,2,1)
sns.histplot(x="MonthlyIncome", hue="Attrition", kde=True,data=df )
plt.title("Employee Attrition by Monthly Income")

plt.subplot(1,2,2)
sns.boxplot(x="Attrition",y="MonthlyIncome",data=df)
plt.title("Employee Attrition by Monthly Income")
plt.tight_layout()
plt.show()


TTest("MonthlyIncome", "Attrition")

(-6.203935765608938, 1.9615812836543436, 7.14736398535381e-10)


p_values = {}
for st in df.select_dtypes("object").columns:
    _,p_values[st],_ = SingleANOVA(st, "MonthlyIncome")


plotP(p_values,tit="p values resulted from ANOVA test between groups based on Monthly income")


SingleANOVA("Department", "MonthlyIncome")

(3.201782929420171, 0.04097409724987449, 3.001858137213299)


sp.posthoc_ttest(df, val_col='MonthlyIncome', group_col='Department').apply(lambda x:round(x, 5))


df['Department'].unique()

array(['Sales', 'Research & Development', 'Human Resources'], dtype=object)


fig1=px.box( df,x="Department" ,y="MonthlyIncome")
fig2=px.box( df,x="Department" ,y="MonthlyIncome", color="Attrition")
fig1.show()
fig2.show()


CalcMEE("Department","MonthlyIncome" ) #95% CL


result= SpecialCalcMEE("Department","MonthlyIncome" )


errors = [float(x[3:]) for x in result["Max Error"]]
px.bar(result, x="Department", y="Mean", error_y= errors, title="Bar Plot of mean Monthly income with max error of estimation")


px.box( df,x="JobLevel" ,y="MonthlyIncome",category_orders= {'JobLevel':order})


CalcMEE("JobLevel","MonthlyIncome" )


result = SpecialCalcMEE("JobLevel","MonthlyIncome" )


errors = [float(x[3:]) for x in result["Max Error"]]
px.bar(result, x="JobLevel", y="Mean", error_y= errors,title="Bar Plot of mean Monthly income with max error of estimation")


sp.posthoc_ttest(df, val_col='MonthlyIncome', group_col='JobRole').apply(lambda x:round(x, 5))


px.box( df,x="JobRole" ,y="MonthlyIncome")


result =SpecialCalcMEE("JobRole","MonthlyIncome" )


errors = [float(x[3:]) for x in result["Max Error"]]
px.bar(result, x="JobRole", y="Mean", error_y= errors,title="Bar Plot of mean Monthly income with max error of estimation")


CalcMEE("JobRole","MonthlyIncome" )


sp.posthoc_ttest(df, val_col='MonthlyIncome', group_col='Education').apply(lambda x:round(x, 5))


px.box( df,x="Education" ,y="MonthlyIncome",points="all")


CalcMEE("Education","MonthlyIncome" )


result =SpecialCalcMEE("Education", "MonthlyIncome")


errors = [float(x[3:]) for x in result["Max Error"]]
px.bar(result, x="Education", y="Mean", error_y= errors,title="Bar Plot of mean Monthly income with max error of estimation")


AttOvered =df.groupby(["OverTime",'Attrition'],as_index=False)['Age'].count()
AttOvered.rename(columns={'Age':'Count'},inplace=True)
AttOvered=AddPercentage(AttOvered,"OverTime")
Overed =df.groupby(["OverTime"],as_index=False)['Age'].count()
Overed.rename(columns={'Age':'Count'},inplace=True)
Overed

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)

(1.6680187953544354, 1.9615812836543436, 0.0955252620565195)

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)

(1.3921381802920636, 1.9615812836543436, 0.16409141231818586)

(14.58158957406234, 2.3213794446188195e-09, 2.6109723453486713)

PearsonRResult(statistic=0.49785456692658037, pvalue=6.669539203000345e-93)

(1.1169671241970975, 3.841458820694124, 0.29057244902890855)

(3.0739613982367193, 9.487729036781154, 0.5455253376565949)


AttOvered =df.groupby(["OverTime",'Attrition'],as_index=False)['Age'].count()
AttOvered.rename(columns={'Age':'Count'},inplace=True)
AttOvered=AddPercentage(AttOvered,"OverTime")
Overed =df.groupby(["OverTime"],as_index=False)['Age'].count()
Overed.rename(columns={'Age':'Count'},inplace=True)
Overed


AttOvered


plt.pie(Overed["Count"], labels=Overed["OverTime"], autopct='%1.1f%%',explode=[0, 0.1])
plt.show()


px.bar(AttOvered,x="OverTime", y="Count", color="Attrition", text="percent")


fig = plt.figure(figsize=(10, 10))
gs = gridspec.GridSpec(1, 2, width_ratios=[1, 1])

ax1 = plt.subplot(gs[0])
ax2 = plt.subplot(gs[1])
ax1.pie(AttOvered[AttOvered["OverTime"]== "Yes"]["Count"], labels=AttOvered[AttOvered["OverTime"]== "Yes"]["Attrition"], autopct='%1.1f%%',explode=[0, 0.1])
ax2.pie(AttOvered[AttOvered["OverTime"]== "No"]["Count"], labels=AttOvered[AttOvered["OverTime"]== "No"]["Attrition"], autopct='%1.1f%%',explode=[0, 0.1])

# Set titles for the subplots
ax1.set_title("Attrition Percentage with Over Time")
ax2.set_title("Attrition Percentage with no Over Time")
ax2.legend()
# Display the pie charts
plt.show()


ChiSquared("OverTime")

(87.56429365828768, 3.841458820694124, 8.15842372153832e-21)


ChiSquared("Department","OverTime")

(0.09360659979986957, 5.991464547107979, 0.9542750851354225)


px.box(df, x="Department", y="MonthlyIncome", color="OverTime")


CalcMEE("OverTime","MonthlyIncome")


TTest("MonthlyIncome","OverTime" )

(0.2333121788024749, 1.9615812836543436, 0.8155515298402164)


px.box(df, x="OverTime", y="MonthlyIncome", points="all")


 def Countdf(st=""):
        dfToPlot=df.groupby([st,'Attrition'])["Age"].count().reset_index(name='Counts')
        dfToPlot["Percentage"]=round((dfToPlot["Counts"]/ sum(dfToPlot["Counts"])) * 100, 2)
        dfToPlot["Percentage"]=dfToPlot["Percentage"].astype(str).apply(lambda x: x +"%")
        return dfToPlot

(1.6680187953544354, 1.9615812836543436, 0.0955252620565195)

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)

(1.3921381802920636, 1.9615812836543436, 0.16409141231818586)

(14.58158957406234, 2.3213794446188195e-09, 2.6109723453486713)

PearsonRResult(statistic=0.49785456692658037, pvalue=6.669539203000345e-93)

(1.1169671241970975, 3.841458820694124, 0.29057244902890855)

(3.0739613982367193, 9.487729036781154, 0.5455253376565949)

(16.024674119585427, 11.070497693516351, 0.006773980139025212)

(46.163676540848705, 5.991464547107979, 9.45551106034083e-11)

(2.994708098265125, 1.9615812836543436, 0.0027930600802134266)


 def Countdf(st=""):
        dfToPlot=df.groupby([st,'Attrition'])["Age"].count().reset_index(name='Counts')
        dfToPlot["Percentage"]=round((dfToPlot["Counts"]/ sum(dfToPlot["Counts"])) * 100, 2)
        dfToPlot["Percentage"]=dfToPlot["Percentage"].astype(str).apply(lambda x: x +"%")
        return dfToPlot


def plotIt(dfToPlot,st=""):
    ax= sns.barplot(
        x = st,
        y = 'Counts',
        hue = 'Attrition',
        data = dfToPlot)
    for i in ax.containers:
        ax.bar_label(i,)


bin_edges = [0, 1, 3, 5, 10]

bin_labels = ['0-1 Companies', '2-3 companies', '4-5 companies', "5+ companies"]
df["NumCompaniesWorkedGroup"] = pd.cut(df['NumCompaniesWorked'], bins=bin_edges, labels=bin_labels)

byNumCompany= Countdf("NumCompaniesWorkedGroup")


plotIt(byNumCompany,"NumCompaniesWorkedGroup")


fig=go.Figure()
byCompany=df.groupby(["NumCompaniesWorkedGroup",'Attrition'],as_index=False)['Age'].count()
byCompany.rename(columns={'Age':'Count'},inplace=True)
fig = make_subplots(rows=1, cols=4,
                        specs=[[{"type": "pie"}, {"type": "pie"},{"type": "pie"}, {"type": "pie"}]],
                        subplot_titles=('0-1 Companies', '2-3 companies','4-5 companies','5+ companies'))

fig.add_trace(go.Pie(values=byCompany[byCompany["NumCompaniesWorkedGroup"]=='0-1 Companies']['Count'],labels=byCompany[byCompany["NumCompaniesWorkedGroup"]=='0-1 Companies']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=1)
fig.add_trace(go.Pie(values=byCompany[byCompany["NumCompaniesWorkedGroup"]=='2-3 companies']['Count'],labels=byCompany[byCompany["NumCompaniesWorkedGroup"]=='2-3 companies']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=2)
fig.add_trace(go.Pie(values=byCompany[byCompany["NumCompaniesWorkedGroup"]=='4-5 companies']['Count'],labels=byCompany[byCompany["NumCompaniesWorkedGroup"]=='4-5 companies']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=3)
fig.add_trace(go.Pie(values=byCompany[byCompany["NumCompaniesWorkedGroup"]=='5+ companies']['Count'],labels=byCompany[byCompany["NumCompaniesWorkedGroup"]=='5+ companies']['Attrition'],pull=[0,0.1],showlegend=True)
                  ,row=1,col=4)

fig.update_layout(title_x=0.5,template='simple_white',showlegend=True,
                      legend_title_text="Attrition",title_text=f"<b style='color:black; font-size:100%;'>Employee Attrition based on number of previos companies",
                      font_family="Times New Roman",title_font_family="Times New Roman")


TTest("NumCompaniesWorked","Attrition" )

(1.6680187953544354, 1.9615812836543436, 0.0955252620565195)


def PlotWithKde(st=""):
    sns.histplot(x=st,hue="Attrition",data=df,kde=True)
    plt.title(f"Employee Distribution by {st}")
    plt.show()


Times =["TotalWorkingYears","YearsAtCompany","YearsSinceLastPromotion","YearsWithCurrManager","YearsInCurrentRole"]
for st in Times:
    PlotWithKde(st)


T_Score = {}
CriticalT = {}
p_values = {}

for col in Times:
    T_Score[col],CriticalT[col],p_values[col] = TTest(col,"Attrition")


columns = list(T_Score.keys())
values = list(T_Score.values())
critical = list(CriticalT.values())

test_df = pd.DataFrame(
    {"Features":columns,
    "T Value":values,
    "Critical Value":critical}
                      )

test_df["P_value"] =  [format(p, '.20f') for p in list(p_values.values())]
test_df["P_value"] = test_df["P_value"].astype(float)
test_df["Result"] = test_df["P_value"].map(lambda x:"Accept" if x > 0.05 else "Reject")
test_df


SatisfactionList = ['EnvironmentSatisfaction', 'JobSatisfaction','RelationshipSatisfaction']

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)

(1.3921381802920636, 1.9615812836543436, 0.16409141231818586)

(14.58158957406234, 2.3213794446188195e-09, 2.6109723453486713)

PearsonRResult(statistic=0.49785456692658037, pvalue=6.669539203000345e-93)

(1.1169671241970975, 3.841458820694124, 0.29057244902890855)

(3.0739613982367193, 9.487729036781154, 0.5455253376565949)

(16.024674119585427, 11.070497693516351, 0.006773980139025212)

(46.163676540848705, 5.991464547107979, 9.45551106034083e-11)

(2.994708098265125, 1.9615812836543436, 0.0027930600802134266)

(16.3250970916474, 7.814727903251179, 0.0009725698845348824)


SatisfactionList = ['EnvironmentSatisfaction', 'JobSatisfaction','RelationshipSatisfaction']


fig=go.Figure()
fig = make_subplots(rows=1, cols=3,
                    specs=[[{"type": "pie"}, {"type": "pie"}, {"type": "pie"}]],
                    subplot_titles=('Environment Satisfaction', 'Job Satisfaction','Relationship Satisfaction'))

EnvironmenSat=df.groupby(["EnvironmentSatisfaction"],as_index=False)["Age"].count()
jobSat=df.groupby(["JobSatisfaction"],as_index=False)["Age"].count()
RelationshipSat=df.groupby(["RelationshipSatisfaction"],as_index=False)["Age"].count()
fig.add_trace(go.Pie(
    values=EnvironmenSat["Age"], 
    labels=EnvironmenSat["EnvironmentSatisfaction"],
    hole=0.5,
    name='Environment Satisfaction',
    showlegend=False)
              ,row=1,col=1)

fig.add_trace(go.Pie(
    values=jobSat["Age"],
    labels=jobSat["JobSatisfaction"],
    hole=0.5,
     name='Job Satisfaction',
    showlegend=False)
              ,row=1,col=2)

fig.add_trace(go.Pie(
    values=RelationshipSat["Age"],
    labels=RelationshipSat["RelationshipSatisfaction"],
    hole=0.5,
    name='Relationship Satisfaction',
    showlegend=True)
              ,row=1,col=3)


fig.update_layout(title_x=0.5,template='simple_white',showlegend=True,
                  legend_title_text="Satisfaction",
                  title_text='<b style="color:black; font-size:100%;">Employee Satisfaction Analysis',
                  font_family="Times New Roman",title_font_family="Times New Roman")


for st in SatisfactionList:
    #print(st)
    #print(ChiSquared(st))
    ChiSquared(st)


for st in CatList[:3]:
    PlotPies(st)


for st in SatisfactionList:
    #print(st)
    #print(SingleANOVA(st,"MonthlyIncome"))
    SingleANOVA(st,"MonthlyIncome")


for st in SatisfactionList:
    #print(st)
    #print(SpecialSingleANOVA(st,"MonthlyIncome"))
    SpecialSingleANOVA(st,"MonthlyIncome")


for st in SatisfactionList:
    fig= px.box(df, x=st, y="MonthlyIncome", color="Attrition")
    fig.show()


sns.kdeplot(x=df['Age'],color='blue',fill=True,label='Age')
plt.axvline(x=df['Age'].mean(),color='red',linestyle ="--",label='Mean Age: 36.923')
plt.legend()
plt.title('Distribution of Age')
plt.show()

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)

(1.3921381802920636, 1.9615812836543436, 0.16409141231818586)

(14.58158957406234, 2.3213794446188195e-09, 2.6109723453486713)

PearsonRResult(statistic=0.49785456692658037, pvalue=6.669539203000345e-93)

(1.1169671241970975, 3.841458820694124, 0.29057244902890855)

(3.0739613982367193, 9.487729036781154, 0.5455253376565949)

(16.024674119585427, 11.070497693516351, 0.006773980139025212)

(46.163676540848705, 5.991464547107979, 9.45551106034083e-11)

(2.994708098265125, 1.9615812836543436, 0.0027930600802134266)

(16.3250970916474, 7.814727903251179, 0.0009725698845348824)


sns.kdeplot(x=df['Age'],color='blue',fill=True,label='Age')
plt.axvline(x=df['Age'].mean(),color='red',linestyle ="--",label='Mean Age: 36.923')
plt.legend()
plt.title('Distribution of Age')
plt.show()


def ShowHist():
    fig, axes = plt.subplots(1, 3, sharex=True, figsize=(15,5))
    fig.suptitle('Attrition Age Distribution by Gender')

    sns.histplot(ax=axes[0],x="Age",hue="Attrition",data=df,kde=True,palette=["r","b"])
    axes[0].set_title('Overall')

    sns.histplot(ax=axes[1],x="Age",hue="Attrition",data=df[df["Gender"]=="Male"].drop([1]),kde=True,palette=["r","b"])
    axes[1].set_title('Male')

    sns.histplot(ax=axes[2],x="Age",hue="Attrition",data=df[df["Gender"]=="Female"],kde=True,palette=["r","b"])
    axes[2].set_title('Female')
    plt.show()


def ShowBox():
    fig, axes = plt.subplots(1, 3, sharex=True, figsize=(15,5))
    fig.suptitle('Attrition Age Distribution by Gender')

    sns.boxplot(ax=axes[0],y="Age",x="Attrition",data=df)

    sns.boxplot(ax=axes[1],y="Age",x="Attrition",data=df[df["Gender"]=="Male"].drop([1]))

    sns.boxplot(ax=axes[2],y="Age",x="Attrition",data=df[df["Gender"]=="Female"])
    plt.show()


TTest("Age", "Attrition")

(-6.178663835307217, 1.9615812836543436, 8.356308021103587e-10)


TTest("Age", "Gender")

(1.3921381802920636, 1.9615812836543436, 0.16409141231818586)


SpecialSingleANOVA("Gender","Age")

(14.58158957406234, 2.3213794446188195e-09, 2.6109723453486713)


px.box(df, x="Gender", y="Age")


ShowHist()
ShowBox()


c= stats.pearsonr(df["MonthlyIncome"], df["Age"])
c

PearsonRResult(statistic=0.49785456692658037, pvalue=6.669539203000345e-93)


slope, intercept, r_value, p_value, std_err = stats.linregress(df['Age'], df['MonthlyIncome'])
line = slope * df['Age'] + intercept

X =np.array(df["Age"])
y=df["MonthlyIncome"]
X_with_intercept = sm.add_constant(X)
SMmodel = sm.OLS(y, X_with_intercept).fit()

# Get predictions and prediction intervals
predictions = SMmodel.get_prediction(X_with_intercept)
pi = predictions.conf_int(obs=True)

max_error =(pi[:,1] - pi[:,0])/2   #(upper - lower)/2    0.95 CI
max_error_df = pd.DataFrame({
    'Age': df['Age'],
    'Upper_bound': line + max_error,
    'Lower_bound': line - max_error
})

fig = px.scatter(df, x='Age', y='MonthlyIncome',trendline='ols',title='Relationship between Age and Monthly Income with Prediction Interval CI = 95%')
fig.update_traces(line=dict(color="rgba(255, 0, 0,0.05)", width=180))

fig.add_trace(px.line(max_error_df, x='Age', y='Upper_bound',line_shape='linear').data[0])
fig.add_trace(px.line(max_error_df, x='Age', y='Lower_bound',line_shape='linear').data[0])
fig.add_trace(px.line( x=df['Age'], y=line,line_shape='linear').data[0])

fig.show()


fig = px.scatter(df, x='Age', y='MonthlyIncome',color="Attrition", trendline='ols'
                 , opacity=0.5,
                 title='Attrition between Age and Monthly Income')

# Show the plot
fig.show()


ChiSquared("Gender", "Attrition")

(1.1169671241970975, 3.841458820694124, 0.29057244902890855)

(3.0739613982367193, 9.487729036781154, 0.5455253376565949)

(16.024674119585427, 11.070497693516351, 0.006773980139025212)

(46.163676540848705, 5.991464547107979, 9.45551106034083e-11)

(2.994708098265125, 1.9615812836543436, 0.0027930600802134266)

(16.3250970916474, 7.814727903251179, 0.0009725698845348824)

<Axes: >

0                       Age
1                 DailyRate
2          DistanceFromHome
4            JobInvolvement
5             MonthlyIncome
9         TotalWorkingYears
10    TrainingTimesLastYear
11           YearsAtCompany
12       YearsInCurrentRole
14     YearsWithCurrManager
Name: Features, dtype: object

Text(0.5, 1.0, 'P values results of correlation analysis between numerical features with Attrition')

<Axes: >


ChiSquared("Gender", "Attrition")

(1.1169671241970975, 3.841458820694124, 0.29057244902890855)


fig = plt.figure(figsize=(10, 5))
gs = gridspec.GridSpec(1, 2, width_ratios=[1, 1])

Gender =df.groupby(["Gender",'Attrition'],as_index=False)['Age'].count()
Gender.rename(columns={'Age':'Counts'},inplace=True)

ax0 = plt.subplot(gs[0])
ax1 = plt.subplot(gs[1])
ax0.pie(Gender[Gender["Gender"]== "Male"]["Counts"], labels=Gender[Gender["Gender"]== "Male"]["Attrition"], autopct='%1.1f%%',explode=[0, 0.1])
ax1.pie(Gender[Gender["Gender"]== "Female"]["Counts"], labels=Gender[Gender["Gender"]== "Female"]["Attrition"], autopct='%1.1f%%',explode=[0, 0.1])

ax0.set_title("Attrition Percentage for Males")
ax1.set_title("Attrition Percentage for Females")

plt.show()


byEduLevel = df.groupby(['Education', 'Attrition'])["Age"].count().reset_index(name='Counts')
byEduLevel = AddPercentage(byEduLevel,"Education")
EduOrder = ["Below College","College","Bachelor","Master","Doctor"]
px.bar(byEduLevel, y ="Counts", x ="Education",color="Attrition", text='percent',category_orders= {'Education':EduOrder})


byEduField = df.groupby(['EducationField', 'Attrition'])["Age"].count().reset_index(name='Counts')
byEduField = AddPercentage(byEduField,"EducationField")
px.bar(byEduField, y ="Counts", x ="EducationField",color="Attrition", text='percent')


_,_ = ANOVA("Education")


ChiSquared("Education")

(3.0739613982367193, 9.487729036781154, 0.5455253376565949)


ChiSquared("EducationField")

(16.024674119585427, 11.070497693516351, 0.006773980139025212)


ChiSquared("MaritalStatus")

(46.163676540848705, 5.991464547107979, 9.45551106034083e-11)

(2.994708098265125, 1.9615812836543436, 0.0027930600802134266)

(16.3250970916474, 7.814727903251179, 0.0009725698845348824)

<Axes: >

0                       Age
1                 DailyRate
2          DistanceFromHome
4            JobInvolvement
5             MonthlyIncome
9         TotalWorkingYears
10    TrainingTimesLastYear
11           YearsAtCompany
12       YearsInCurrentRole
14     YearsWithCurrManager
Name: Features, dtype: object

Text(0.5, 1.0, 'P values results of correlation analysis between numerical features with Attrition')

<Axes: >

3                  HourlyRate
6                 MonthlyRate
7          NumCompaniesWorked
8           PercentSalaryHike
13    YearsSinceLastPromotion
Name: Features, dtype: object

{'BusinessTravel',
 'Department',
 'EducationField',
 'EnvironmentSatisfaction',
 'JobLevel',
 'JobRole',
 'JobSatisfaction',
 'MaritalStatus',
 'OverTime',
 'StockOptionLevel',
 'WorkLifeBalance'}

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])


ChiSquared("MaritalStatus")

(46.163676540848705, 5.991464547107979, 9.45551106034083e-11)


bus=df.groupby(["MaritalStatus",'Attrition'],as_index=False)['Age'].count()
bus.rename(columns={'Age':'Count'},inplace=True)
fig=go.Figure()
fig = make_subplots(rows=1, cols=3,
                        specs=[[{"type": "pie"}, {"type": "pie"},{"type": "pie"}]],
                        subplot_titles=('Divorced', 'Married','Single'))

fig.add_trace(go.Pie(values=bus[bus["MaritalStatus"]=='Divorced']['Count'],labels=bus[bus["MaritalStatus"]=='Divorced']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=1)
fig.add_trace(go.Pie(values=bus[bus["MaritalStatus"]=='Married']['Count'],labels=bus[bus["MaritalStatus"]=='Married']['Attrition'],pull=[0,0.1],showlegend=False)
                  ,row=1,col=2)
fig.add_trace(go.Pie(values=bus[bus["MaritalStatus"]=='Single']['Count'],labels=bus[bus["MaritalStatus"]=='Single']['Attrition'],pull=[0,0.1],showlegend=True)
                  ,row=1,col=3)
   
fig.update_layout(title_x=0.5,template='simple_white',showlegend=True,
                      legend_title_text="Attrition",title_text=f"<b style='color:black; font-size:100%;'>Employee Attrition based on Marital Status",
                      font_family="Times New Roman",title_font_family="Times New Roman")
fig.show()


px.box(df, y="DistanceFromHome",x="Attrition", color="Attrition", points="all" )


plt.figure(figsize=(13.5,6))
plt.subplot(1,2,1)
sns.histplot(x="DistanceFromHome",hue="Attrition",data=df,kde=True)
plt.title("Employee Distribution by Distance From Home & Attrition")

plt.subplot(1,2,2)
sns.boxplot(x="Attrition",y="DistanceFromHome",data=df)
plt.title("Employee Distribution by Distance From Home & Attrition")

plt.show()


TTest("DistanceFromHome", "Attrition")

(2.994708098265125, 1.9615812836543436, 0.0027930600802134266)


WL_Balance=df.groupby(["WorkLifeBalance"],as_index=False)["Age"].count()
fig =px.pie(
    values=WL_Balance["Age"],
   names=WL_Balance["WorkLifeBalance"],title="Work life balance rate in the orgnization",
    hole=0.5)
fig.show()


PlotPies("WorkLifeBalance")


ChiSquared("WorkLifeBalance")

(16.3250970916474, 7.814727903251179, 0.0009725698845348824)


corr = df.select_dtypes("number").corr()
mask = np.zeros_like(corr)
mask[np.triu_indices_from(mask)] = True
plt.figure(figsize=(15, 10))
sns.heatmap(corr,
            vmax=1,
            mask=mask,
            annot=True, fmt='.2f',
            linewidths=.2, cmap=sns.color_palette("Reds"))

<Axes: >

0                       Age
1                 DailyRate
2          DistanceFromHome
4            JobInvolvement
5             MonthlyIncome
9         TotalWorkingYears
10    TrainingTimesLastYear
11           YearsAtCompany
12       YearsInCurrentRole
14     YearsWithCurrManager
Name: Features, dtype: object

Text(0.5, 1.0, 'P values results of correlation analysis between numerical features with Attrition')

<Axes: >

3                  HourlyRate
6                 MonthlyRate
7          NumCompaniesWorked
8           PercentSalaryHike
13    YearsSinceLastPromotion
Name: Features, dtype: object

{'BusinessTravel',
 'Department',
 'EducationField',
 'EnvironmentSatisfaction',
 'JobLevel',
 'JobRole',
 'JobSatisfaction',
 'MaritalStatus',
 'OverTime',
 'StockOptionLevel',
 'WorkLifeBalance'}

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])

OneHotEncoder(cols=['BusinessTravel', 'Department', 'EducationField',
                    'EnvironmentSatisfaction', 'JobLevel', 'JobRole',
                    'JobSatisfaction', 'MaritalStatus', 'OverTime',
                    'RelationshipSatisfaction', 'StockOptionLevel',
                    'WorkLifeBalance'],
              use_cat_names=True)

StandardScaler()


corr = df.select_dtypes("number").corr()
mask = np.zeros_like(corr)
mask[np.triu_indices_from(mask)] = True
plt.figure(figsize=(15, 10))
sns.heatmap(corr,
            vmax=1,
            mask=mask,
            annot=True, fmt='.2f',
            linewidths=.2, cmap=sns.color_palette("Reds"))

<Axes: >


CorrScore = {}
p_values = {}
for col in df.select_dtypes("number").columns:
    cc, p = stats.pointbiserialr(df["Attrition"].replace({"Yes":1,"No": 0}), df[col])
    CorrScore[col] = cc
    p_values[col] = p

0                       Age
1                 DailyRate
2          DistanceFromHome
4            JobInvolvement
5             MonthlyIncome
9         TotalWorkingYears
10    TrainingTimesLastYear
11           YearsAtCompany
12       YearsInCurrentRole
14     YearsWithCurrManager
Name: Features, dtype: object

Text(0.5, 1.0, 'P values results of correlation analysis between numerical features with Attrition')

<Axes: >

3                  HourlyRate
6                 MonthlyRate
7          NumCompaniesWorked
8           PercentSalaryHike
13    YearsSinceLastPromotion
Name: Features, dtype: object

{'BusinessTravel',
 'Department',
 'EducationField',
 'EnvironmentSatisfaction',
 'JobLevel',
 'JobRole',
 'JobSatisfaction',
 'MaritalStatus',
 'OverTime',
 'StockOptionLevel',
 'WorkLifeBalance'}

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])

OneHotEncoder(cols=['BusinessTravel', 'Department', 'EducationField',
                    'EnvironmentSatisfaction', 'JobLevel', 'JobRole',
                    'JobSatisfaction', 'MaritalStatus', 'OverTime',
                    'RelationshipSatisfaction', 'StockOptionLevel',
                    'WorkLifeBalance'],
              use_cat_names=True)

StandardScaler()

LogisticRegression(max_iter=1000)


CorrScore = {}
p_values = {}
for col in df.select_dtypes("number").columns:
    cc, p = stats.pointbiserialr(df["Attrition"].replace({"Yes":1,"No": 0}), df[col])
    CorrScore[col] = cc
    p_values[col] = p


columns = list(CorrScore.keys())
values = list(CorrScore.values())

test_df = pd.DataFrame(
    {"Features":columns,
    "Corr":values}
                      )
test_df["P_value"] =  [format(p, '.20f') for p in list(p_values.values())]
test_df["P_value"] = test_df["P_value"].astype(float)
test_df["Result"] = test_df["P_value"].map(lambda x:"Accept" if x > 0.05 else "Reject")
test_df


test_df[test_df["P_value"]<0.05]["Features"]

0                       Age
1                 DailyRate
2          DistanceFromHome
4            JobInvolvement
5             MonthlyIncome
9         TotalWorkingYears
10    TrainingTimesLastYear
11           YearsAtCompany
12       YearsInCurrentRole
14     YearsWithCurrManager
Name: Features, dtype: object

[
 "Age"                   , "DailyRate"
 "DistanceFromHome"      , "JobInvolvement"
 "MonthlyIncome"         , "TotalWorkingYears"
 "TrainingTimesLastYear" , "YearsAtCompany"
 "YearsInCurrentRole"    , "YearsWithCurrManager"
]


PlotErrors(p_values)
plt.title("P values results of correlation analysis between numerical features with Attrition")

Text(0.5, 1.0, 'P values results of correlation analysis between numerical features with Attrition')


df.select_dtypes("number").corrwith(df["Attrition"].replace({"Yes":1,"No": 0})).sort_values().plot(kind='barh')

<Axes: >


test_df, p= ApplyTTest("Attrition")
test_df


plotP(p,"")


test_df[test_df["P_value"] > 0.05]["Features"]

3                  HourlyRate
6                 MonthlyRate
7          NumCompaniesWorked
8           PercentSalaryHike
13    YearsSinceLastPromotion
Name: Features, dtype: object

[
 "Age"                   , "DailyRate"
 "DistanceFromHome"      , "JobInvolvement"
 "MonthlyIncome"         , "TotalWorkingYears"
 "TrainingTimesLastYear" , "YearsAtCompany"
 "YearsInCurrentRole"    , "YearsWithCurrManager"
 "NumCompaniesWorked"
]


chi2_statistic = {}
p_values = {}
criticalChis = {} 
# Perform chi-square test for each column
for col in df.select_dtypes("object").drop("Attrition", axis =1).columns:
    chi2, crit,p_value = ChiSquared(col)
    chi2_statistic[col] = chi2
    p_values[col] = p_value
    criticalChis[col] = crit


columns = list(chi2_statistic.keys())
values = list(chi2_statistic.values())
critical = list(criticalChis.values())

test_df = pd.DataFrame(
    {"Features":columns,
    "Chi_2 Statistic":values,
    "Critical Value":critical}
                      )
test_df["P_value"] =  [format(p, '.20f') for p in list(p_values.values())]
test_df["P_value"] = test_df["P_value"].astype(float)
test_df["Result"] = test_df["P_value"].map(lambda x:"Accept" if x > 0.05 else "Reject")
test_df


set(test_df[test_df["P_value"]<0.05]["Features"])

{'BusinessTravel',
 'Department',
 'EducationField',
 'EnvironmentSatisfaction',
 'JobLevel',
 'JobRole',
 'JobSatisfaction',
 'MaritalStatus',
 'OverTime',
 'StockOptionLevel',
 'WorkLifeBalance'}

[
 "BusinessTravel"   , "Department"
 "EducationField"   , "EnvironmentSatisfaction"
 "JobLevel"         , "JobRole"
 "JobSatisfaction"  , "MaritalStatus"
 "OverTime"         , "StockOptionLevel"
 "WorkLifeBalance"
]


PlotChi(columns,values)


plotP(p_values,"",True)


PlotErrors(p_values,ISchi = True)


def evaluate(model, X_train, X_test, y_train, y_test):
    y_test_pred = model.predict(X_test)
    y_train_pred = model.predict(X_train)
    acc_train = accuracy_score(y_train, y_train_pred)
    acc_test = accuracy_score(y_test,y_test_pred)
    testconfusion = confusion_matrix(y_test, y_test_pred)
    trainconfusion = confusion_matrix(y_train, y_train_pred)

    print("TRAINIG RESULTS: \n===============================")
    
    print(f"CONFUSION MATRIX:\n{trainconfusion}")
    print(f"ACCURACY SCORE:\n{acc_train:.4f}")
    print("precision score:", round(precision_score(y_train,y_train_pred),2))
    print("Recall Accuracy:", round(recall_score(y_train,y_train_pred),2))
    print("Area Under Curve AUC:", round(roc_auc_score(y_train,y_train_pred),2))
    
    print("\n\nTRAINIG RESULTS: \n===============================")
    
    print(f"CONFUSION MATRIX:\n{testconfusion}")
    print(f"ACCURACY SCORE:\n{acc_test:.4f}")
    print("precision score:", round(precision_score(y_test,y_test_pred),2))
    print("Recall Accuracy:", round(recall_score(y_test,y_test_pred),2))
    print("Area Under Curve AUC:", round(roc_auc_score(y_test,y_test_pred),2))


data =df.copy()
data.drop(columns="NumCompaniesWorkedGroup", inplace =True)


data["Attrition"]= data["Attrition"].replace({"Yes":1,
"No": 0})


from category_encoders import OneHotEncoder


X = data.drop(columns=['Attrition','HourlyRate','MonthlyRate',
                       'NumCompaniesWorked','PercentSalaryHike','YearsSinceLastPromotion',
                      'JobInvolvement','Education','Gender','YearsAtCompany','PerformanceRating','YearsWithCurrManager'], axis=1)
y = data.Attrition

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42, 
                                                    stratify=y)# because the data is unbalanced
model = make_pipeline(
    OneHotEncoder(use_cat_names=True),
    StandardScaler(),
    LogisticRegression(max_iter=1000)
)
model.fit(X_train,y_train)

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['BusinessTravel', 'Department',
                                     'EducationField',
                                     'EnvironmentSatisfaction', 'JobLevel',
                                     'JobRole', 'JobSatisfaction',
                                     'MaritalStatus', 'OverTime',
                                     'RelationshipSatisfaction',
                                     'StockOptionLevel', 'WorkLifeBalance'],
                               use_cat_names=True)),
                ('standardscaler', StandardScaler()),
                ('logisticregression', LogisticRegression(max_iter=1000))])

OneHotEncoder(cols=['BusinessTravel', 'Department', 'EducationField',
                    'EnvironmentSatisfaction', 'JobLevel', 'JobRole',
                    'JobSatisfaction', 'MaritalStatus', 'OverTime',
                    'RelationshipSatisfaction', 'StockOptionLevel',
                    'WorkLifeBalance'],
              use_cat_names=True)

StandardScaler()

LogisticRegression(max_iter=1000)


features = model.named_steps["onehotencoder"].get_feature_names_out()
importances = model.named_steps["logisticregression"].coef_[0]


evaluate(model, X_train, X_test, y_train, y_test)

TRAINIG RESULTS: 
===============================
CONFUSION MATRIX:
[[843  20]
 [ 83  83]]
ACCURACY SCORE:
0.8999
precision score: 0.81
Recall Accuracy: 0.5
Area Under Curve AUC: 0.74


TRAINIG RESULTS: 
===============================
CONFUSION MATRIX:
[[357  13]
 [ 42  29]]
ACCURACY SCORE:
0.8753
precision score: 0.69
Recall Accuracy: 0.41
Area Under Curve AUC: 0.69


ConfusionMatrixDisplay.from_estimator(model, X_train, y_train)

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x181699ee250>


ConfusionMatrixDisplay.from_estimator(model, X_test, y_test)

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x18169c88490>


numOfSumples = 50


Accuracies = list()
for i in range(numOfSumples):
    BootX_train , Booty_train = resample(X_train, y_train)
    model.fit(BootX_train, Booty_train)
    y_test_pred = model.predict(X_test)
    score =accuracy_score(y_test, y_test_pred)
    #print("TESTING RESULTS: \n===============================")
    #print(f"CONFUSION MATRIX:\n{confusion_matrix(y_test, y_test_pred)}")
    #print(f"ACCURACY SCORE:\n{score:.4f}")
    Accuracies.append(score)


plt.errorbar(x = np.arange(1, len(Accuracies) +1 ,1), y= Accuracies, fmt='o')
plt.axhline(y=np.mean(Accuracies),color='red',linestyle ="--", label=f"Mean Accuracy = {round(np.mean(Accuracies),4)}")
plt.legend()

<matplotlib.legend.Legend at 0x1816a6eb810>


sns.histplot(Accuracies, kde= True)

<Axes: ylabel='Count'>


# confidence intervals
alpha = 0.95
p = ((1.0-alpha)/2.0) * 100
lower = max(0.0, np.percentile(Accuracies, p))
p = (alpha+((1.0-alpha)/2.0)) * 100
upper = min(1.0, np.percentile(Accuracies, p))


plt.figure(figsize=(12,6))

x= np.arange(1, len(Accuracies) +1 ,1)
y= Accuracies
y_error =(upper - lower) / 2.0
 
plt.errorbar(x, y,
             yerr = y_error,
             fmt ='o',capsize=3)
plt.xticks(rotation=90)

plt.axhline(y=np.mean(Accuracies),color='red')

<matplotlib.lines.Line2D at 0x18169ddc510>


print('At %.1f confidence interval Accuracy =  %.1f%% +/- %.1f%%' % (alpha*100,np.mean(Accuracies)*100, (upper -lower)/2.0*100))

At 95.0 confidence interval Accuracy =  86.0% +/- 1.7%


print('At %.1f confidence interval Accuracy lies between %.1f%% and %.1f%%' % (alpha*100, lower*100, upper*100))

At 95.0 confidence interval Accuracy lies between 84.1% and 87.5%


Resuls = df.copy()


Resuls["AttrionLikelihood"] = model.predict_proba(X)[:, 1]


sns.histplot(Resuls["AttrionLikelihood"], kde=True)

<Axes: xlabel='AttrionLikelihood', ylabel='Count'>


Resuls["RiskLevel"] = Resuls["AttrionLikelihood"].astype("object").map(lambda x:"Strong" if x > 0.4 else "Medium" if x > 0.3 else "Weak" )


Resuls[['Attrition', 'AttrionLikelihood', 'RiskLevel' ]][42:53]


pd.crosstab(Resuls["RiskLevel"] ,Resuls["Attrition"])


feat_imp = pd.Series(importances, index=features).sort_values()
SortedDF= pd.DataFrame(feat_imp)
MaxMin =pd.concat([SortedDF.head(7), SortedDF.tail(7)], axis=0)

MaxMin.plot(kind="barh")
plt.xlabel("Importance")
plt.ylabel("Feature")
plt.title("Feature Importance")

Text(0.5, 1.0, 'Feature Importance')


def make_prediction(age, businesstravel,dailyrate,department,distanceFromHome,
                    educationfield,environmentsatisfaction,
                    joblevel,jobrole,jobsatisfaction,maritalstatus,
                    monthlyincome,overtime,relationshipsatisfaction,
                    stockoptionlevel,totalworkingyears,trainingtimeslastyear,
                    worklifebalance,yearsincurrentrole
                   ):
    data={
        'Age':age,
        'BusinessTravel':businesstravel,
        'DailyRate':dailyrate,
         'Department':department,
        'DistanceFromHome' :distanceFromHome,
       'EducationField':educationfield,
        'EnvironmentSatisfaction':environmentsatisfaction,
        'JobLevel':joblevel,
        'JobRole':jobrole,
        'JobSatisfaction':jobsatisfaction,
        'MaritalStatus':maritalstatus,
       'MonthlyIncome':monthlyincome,
        'OverTime':overtime,
        'RelationshipSatisfaction':relationshipsatisfaction,
       'StockOptionLevel':stockoptionlevel,
        'TotalWorkingYears':totalworkingyears,
        'TrainingTimesLastYear':trainingtimeslastyear,
        'WorkLifeBalance':worklifebalance,
         'YearsInCurrentRole':yearsincurrentrole
    }
    df=pd.DataFrame(data,index=[0])
    prediction = model.predict_proba(df)[:, 1][0]
    if prediction > 0.4:
        Risk ="Strong"
    elif prediction > 0.3:
        Risk = "Medium"
    else:
        Risk ="Weak" 
    return f"This employee has a {Risk} Risk to quit with Probability = {round(prediction,5)}."


interact(
    make_prediction,
    age=IntSlider(min=X_train["Age"].min(), max=X_train["Age"].max(),value=X_train["Age"].mean()),
    businesstravel=Dropdown(options=sorted(X_train["BusinessTravel"].unique())),
    dailyrate=IntSlider(min=X_train["DailyRate"].min(), max=X_train["DailyRate"].max(),value=X_train["DailyRate"].mean()),    
    department=Dropdown(options=sorted(X_train["Department"].unique())),
    distanceFromHome=IntSlider(min=X_train["DistanceFromHome"].min(), max=X_train["DistanceFromHome"].max(),value=X_train["DistanceFromHome"].mean()),
    educationfield=Dropdown(options=sorted(X_train["EducationField"].unique())),
    environmentsatisfaction=Dropdown(options=sorted(X_train["EnvironmentSatisfaction"].unique())),
    joblevel=Dropdown(options=sorted(X_train["JobLevel"].unique())),
    jobrole=Dropdown(options=sorted(X_train["JobRole"].unique())),
    jobsatisfaction=Dropdown(options=sorted(X_train["JobSatisfaction"].unique())),
    maritalstatus=Dropdown(options=sorted(X_train["MaritalStatus"].unique())),
    monthlyincome=IntSlider(min=X_train["MonthlyIncome"].min(), max=X_train["MonthlyIncome"].max(),value=X_train["MonthlyIncome"].mean()),
    overtime=Dropdown(options=sorted(X_train["OverTime"].unique())),
    relationshipsatisfaction=Dropdown(options=sorted(X_train["RelationshipSatisfaction"].unique())),
    stockoptionlevel=Dropdown(options=sorted(X_train["StockOptionLevel"].unique())),
    totalworkingyears=IntSlider(min=X_train["TotalWorkingYears"].min(), max=X_train["TotalWorkingYears"].max(),value=X_train["TotalWorkingYears"].mean()),
    trainingtimeslastyear=IntSlider(min=X_train["TrainingTimesLastYear"].min(), max=X_train["TrainingTimesLastYear"].max(),value=X_train["TrainingTimesLastYear"].mean()),
    worklifebalance=Dropdown(options=sorted(X_train["WorkLifeBalance"].unique())),
    yearsincurrentrole=IntSlider(min=X_train["YearsInCurrentRole"].min(), max=X_train["YearsInCurrentRole"].max(),value=X_train["YearsInCurrentRole"].mean()),

);

interactive(children=(IntSlider(value=36, description='age', max=60, min=18), Dropdown(description='businesstr…

	Age	DailyRate	DistanceFromHome	EmployeeCount	EmployeeNumber	HourlyRate	JobInvolvement	MonthlyIncome	MonthlyRate	NumCompaniesWorked	PercentSalaryHike	StandardHours	TotalWorkingYears	TrainingTimesLastYear	YearsAtCompany	YearsInCurrentRole	YearsSinceLastPromotion	YearsWithCurrManager
count	1470.000000	1470.000000	1470.000000	1470.0	1470.000000	1470.000000	1470.000000	1470.000000	1470.000000	1470.000000	1470.000000	1470.0	1470.000000	1470.000000	1470.000000	1470.000000	1470.000000	1470.000000
mean	36.923810	802.485714	9.192517	1.0	1024.865306	65.891156	2.729932	6502.931293	14313.103401	2.693197	15.209524	80.0	11.279592	2.799320	7.008163	4.229252	2.187755	4.123129
std	9.135373	403.509100	8.106864	0.0	602.024335	20.329428	0.711561	4707.956783	7117.786044	2.498009	3.659938	0.0	7.780782	1.289271	6.126525	3.623137	3.222430	3.568136
min	18.000000	102.000000	1.000000	1.0	1.000000	30.000000	1.000000	1009.000000	2094.000000	0.000000	11.000000	80.0	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	30.000000	465.000000	2.000000	1.0	491.250000	48.000000	2.000000	2911.000000	8047.000000	1.000000	12.000000	80.0	6.000000	2.000000	3.000000	2.000000	0.000000	2.000000
50%	36.000000	802.000000	7.000000	1.0	1020.500000	66.000000	3.000000	4919.000000	14235.500000	2.000000	14.000000	80.0	10.000000	3.000000	5.000000	3.000000	1.000000	3.000000
75%	43.000000	1157.000000	14.000000	1.0	1555.750000	83.750000	3.000000	8379.000000	20461.500000	4.000000	18.000000	80.0	15.000000	3.000000	9.000000	7.000000	3.000000	7.000000
max	60.000000	1499.000000	29.000000	1.0	2068.000000	100.000000	4.000000	19999.000000	26999.000000	9.000000	25.000000	80.0	40.000000	6.000000	40.000000	18.000000	15.000000	17.000000

	Sales	Research & Development	Human Resources
Sales	1.000000	0.010998	0.599484
Research & Development	0.010998	1.000000	0.562536
Human Resources	0.599484	0.562536	1.000000

	Sales	Research & Development	Human Resources
Sales	1.00000	0.01100	0.59948
Research & Development	0.01100	1.00000	0.56254
Human Resources	0.59948	0.56254	1.00000

	JobLevel	Total number	Standard Deviation	Mean	Max Error
0	Junior Level	513	1161.309722	5334.621832	+/- 100.7315
1	Entry Level	526	665.460059	2721.863118	+/- 57.0006
2	Mid Level	218	1805.999233	9817.252294	+/- 241.0828
3	Senior Level	106	1816.239003	15503.783019	+/- 349.7859
4	Executive Level	69	512.383127	19191.826087	+/- 123.0879

	Sales Executive	Research Scientist	Laboratory Technician	Manufacturing Director	Healthcare Representative	Manager	Sales Representative	Research Director	Human Resources
Sales Executive	1.00000	0.00000	0.00000	0.13262	0.01607	0.00000	0.00000	0.00000	0.00000
Research Scientist	0.00000	1.00000	0.97773	0.00000	0.00000	0.00000	0.00002	0.00000	0.00001
Laboratory Technician	0.00000	0.97773	1.00000	0.00000	0.00000	0.00000	0.00001	0.00000	0.00001
Manufacturing Director	0.13262	0.00000	0.00000	1.00000	0.45905	0.00000	0.00000	0.00000	0.00000
Healthcare Representative	0.01607	0.00000	0.00000	0.45905	1.00000	0.00000	0.00000	0.00000	0.00000
Manager	0.00000	0.00000	0.00000	0.00000	0.00000	1.00000	0.00000	0.00298	0.00000
Sales Representative	0.00000	0.00002	0.00001	0.00000	0.00000	0.00000	1.00000	0.00000	0.00000
Research Director	0.00000	0.00000	0.00000	0.00000	0.00000	0.00298	0.00000	1.00000	0.00000
Human Resources	0.00000	0.00001	0.00001	0.00000	0.00000	0.00000	0.00000	0.00000	1.00000

	Age	Attrition	BusinessTravel	DailyRate	Department	DistanceFromHome	Education	EducationField	EmployeeCount	EmployeeNumber	...	RelationshipSatisfaction	StandardHours	StockOptionLevel	TotalWorkingYears	TrainingTimesLastYear	WorkLifeBalance	YearsAtCompany	YearsInCurrentRole	YearsSinceLastPromotion	YearsWithCurrManager
0	41	Yes	Travel_Rarely	1102	Sales	1	2	Life Sciences	1	1	...	1	80	0	8	0	1	6	4	0	5
1	49	No	Travel_Frequently	279	Research & Development	8	1	Life Sciences	1	2	...	4	80	1	10	3	3	10	7	1	7
2	37	Yes	Travel_Rarely	1373	Research & Development	2	2	Other	1	4	...	2	80	0	7	3	3	0	0	0	0
3	33	No	Travel_Frequently	1392	Research & Development	3	4	Life Sciences	1	5	...	3	80	0	8	3	3	8	7	3	0
4	27	No	Travel_Rarely	591	Research & Development	2	1	Medical	1	7	...	4	80	1	6	3	3	2	2	2	2

	Unique	Type
Age	43	int64
Attrition	2	object
BusinessTravel	3	object
DailyRate	886	int64
Department	3	object
DistanceFromHome	29	int64
Education	5	object
EducationField	6	object
EmployeeCount	1	int64
EmployeeNumber	1470	int64
EnvironmentSatisfaction	4	object
Gender	2	object
HourlyRate	71	int64
JobInvolvement	4	int64
JobLevel	5	object
JobRole	9	object
JobSatisfaction	4	object
MaritalStatus	3	object
MonthlyIncome	1349	int64
MonthlyRate	1427	int64
NumCompaniesWorked	10	int64
Over18	1	object
OverTime	2	object
PercentSalaryHike	15	int64
PerformanceRating	2	object
RelationshipSatisfaction	4	object
StandardHours	1	int64
StockOptionLevel	4	object
TotalWorkingYears	40	int64
TrainingTimesLastYear	7	int64
WorkLifeBalance	4	object
YearsAtCompany	37	int64
YearsInCurrentRole	19	int64
YearsSinceLastPromotion	16	int64
YearsWithCurrManager	18	int64

	Department	Total number	Standard Deviation	Mean	Max Error
0	Sales	415	2867.500943	6153.732530	+/- 276.6936
1	Research & Development	869	3146.772819	5027.997699	+/- 209.5123
2	Human Resources	55	3559.526833	4864.018182	+/- 962.2749

	JobRole	Total number	Standard Deviation	Mean	Max Error
0	Sales Executive	325	2338.860456	6902.901538	+/- 255.2325
1	Research Scientist	285	1036.962913	3146.663158	+/- 120.9048
2	Laboratory Technician	254	1049.060730	3168.397638	+/- 129.6326
3	Manufacturing Director	145	2676.745753	7295.137931	+/- 439.3761
4	Healthcare Representative	131	2542.550170	7528.763359	+/- 439.4846
5	Manager	94	1751.984116	17644.212766	+/- 358.8411
6	Sales Representative	72	447.921852	2550.986111	+/- 105.2565
7	Research Director	80	2827.621369	16033.550000	+/- 629.2563
8	Human Resources	52	2438.849744	4235.750000	+/- 678.9801

	College	Below College	Master	Bachelor	Doctor
College	1.00000	0.18144	0.09133	0.39796	0.00462
Below College	0.18144	1.00000	0.00491	0.03469	0.00057
Master	0.09133	0.00491	1.00000	0.30994	0.04484
Bachelor	0.39796	0.03469	0.30994	1.00000	0.01571
Doctor	0.00462	0.00057	0.04484	0.01571	1.00000

	Education	Total number	Standard Deviation	Mean	Max Error
0	College	251	2368.879924	4866.195219	+/- 294.4841
1	Below College	159	3113.741827	4776.276730	+/- 487.7209
2	Master	368	3336.802687	5873.663043	+/- 342.0498
3	Bachelor	544	3966.447131	5866.963235	+/- 334.0561
4	Doctor	48	5061.430495	8277.645833	+/- 1469.6862

	OverTime	Attrition	Count	percent
0	No	No	944	89.56 %
1	No	Yes	110	10.44 %
2	Yes	No	289	69.47 %
3	Yes	Yes	127	30.53 %

	OverTime	Total number	Standard Deviation	Mean	Max Error
0	Yes	388	3502.004143	5685.396907	+/- 349.5499
1	No	969	3261.547377	5441.221878	+/- 205.6143

	Features	T Value	Critical Value	P_value	Result
0	TotalWorkingYears	-6.652255	1.961581	4.061878e-11	Reject
1	YearsAtCompany	-5.196309	1.961581	2.318872e-07	Reject
2	YearsSinceLastPromotion	-1.265788	1.961581	2.057900e-01	Accept
3	YearsWithCurrManager	-6.059069	1.961581	1.736987e-09	Reject
4	YearsInCurrentRole	-6.232038	1.961581	6.003186e-10	Reject

Satisfaction	${X^2}$	${X^2}_c$	$Pvalue$
EnvironmentSatisfaction	22.5038	7.8147	5.1234e-05
JobSatisfaction	17.5051	7.8147	0.0005
RelationshipSatisfaction	5.2411	7.8147	0.1549

Satisfaction	$F$	$F_c$	$Pvalue$
EnvironmentSatisfaction	0.4100	2.6109	0.7458
JobSatisfaction	0.0270	2.6109	0.9940
RelationshipSatisfaction	0.5529	2.6109	0.6462

Satisfaction	$F$	$F_c$	$Pvalue$
EnvironmentSatisfaction	6.2719	2.0158	2.8786 $*{10}^{-7}$
JobSatisfaction	5.6935	2.0158	1.6477 $*{10}^{-6}$
RelationshipSatisfaction	6.5153	2.0158	1.3754 $*{10}^{-7}$

	Features	Chi_2 Statistic	Critical Value	P_value	Result
0	BusinessTravel	24.182414	5.991465	5.608614e-06	Reject
1	Department	10.796007	5.991465	4.525607e-03	Reject
2	Education	3.073961	9.487729	5.455253e-01	Accept
3	EducationField	16.024674	11.070498	6.773980e-03	Reject
4	EnvironmentSatisfaction	22.503881	7.814728	5.123469e-05	Reject
5	Gender	1.116967	3.841459	2.905724e-01	Accept
6	JobLevel	72.529013	9.487729	6.634680e-15	Reject
7	JobRole	86.190254	15.507313	2.752480e-15	Reject
8	JobSatisfaction	17.505077	7.814728	5.563005e-04	Reject
9	MaritalStatus	46.163677	5.991465	9.455511e-11	Reject
10	OverTime	87.564294	3.841459	1.000000e-20	Reject
11	PerformanceRating	0.000155	3.841459	9.900745e-01	Accept
12	RelationshipSatisfaction	5.241068	7.814728	1.549724e-01	Accept
13	StockOptionLevel	60.598301	7.814728	4.379390e-13	Reject
14	WorkLifeBalance	16.325097	7.814728	9.725699e-04	Reject

	Attrition	AttrionLikelihood	RiskLevel
42	Yes	0.811160	Strong
43	No	0.288771	Weak
44	No	0.024744	Weak
45	Yes	0.000037	Weak
46	No	0.012220	Weak
47	No	0.208084	Weak
48	No	0.213677	Weak
49	No	0.069798	Weak
50	Yes	0.784926	Strong
51	Yes	0.762197	Strong
52	No	0.095868	Weak

	Features	Corr	P_value	Result
0	Age	-0.159205	8.356308e-10	Reject
1	DailyRate	-0.056652	2.985816e-02	Reject
2	DistanceFromHome	0.077924	2.793060e-03	Reject
3	HourlyRate	-0.006846	7.931348e-01	Accept
4	JobInvolvement	-0.130016	5.677065e-07	Reject
5	MonthlyIncome	-0.159840	7.147364e-10	Reject
6	MonthlyRate	0.015170	5.611236e-01	Accept
7	NumCompaniesWorked	0.043494	9.552526e-02	Accept
8	PercentSalaryHike	-0.013478	6.056128e-01	Accept
9	TotalWorkingYears	-0.171063	4.061878e-11	Reject
10	TrainingTimesLastYear	-0.059478	2.257850e-02	Reject
11	YearsAtCompany	-0.134392	2.318872e-07	Reject
12	YearsInCurrentRole	-0.160545	6.003186e-10	Reject
13	YearsSinceLastPromotion	-0.033019	2.057900e-01	Accept
14	YearsWithCurrManager	-0.156199	1.736987e-09	Reject

	Features	T Value	Critical Value	P_value	Result
0	Age	-6.178664	1.961581	8.356308e-10	Reject
1	DailyRate	-2.174084	1.961581	2.985816e-02	Reject
2	DistanceFromHome	2.994708	1.961581	2.793060e-03	Reject
3	HourlyRate	-0.262290	1.961581	7.931348e-01	Accept
4	JobInvolvement	-5.024140	1.961581	5.677065e-07	Reject
5	MonthlyIncome	-6.203936	1.961581	7.147364e-10	Reject
6	MonthlyRate	0.581306	1.961581	5.611236e-01	Accept
7	NumCompaniesWorked	1.668019	1.961581	9.552526e-02	Accept
8	PercentSalaryHike	-0.516457	1.961581	6.056128e-01	Accept
9	TotalWorkingYears	-6.652255	1.961581	4.061878e-11	Reject
10	TrainingTimesLastYear	-2.282903	1.961581	2.257850e-02	Reject
11	YearsAtCompany	-5.196309	1.961581	2.318872e-07	Reject
12	YearsInCurrentRole	-6.232038	1.961581	6.003186e-10	Reject
13	YearsSinceLastPromotion	-1.265788	1.961581	2.057900e-01	Accept
14	YearsWithCurrManager	-6.059069	1.961581	1.736987e-09	Reject

Welcome to our project that focused on predicting employee turnover and set a risk level for each employee.

Table of Contants¶

Let's rename some categorical values to a meaningful names¶

Our methodology to analyse features:¶

overview ¶

Amazing¶

No duplicate, no NULL¶

What is the overall attrition rate in the organization?¶

Catalog of used functions' implementations:¶

Considering Technical reasons ¶

Department Feature:¶

Is Departement has an effect on the decision of quit ?¶

Is there a significant difference among groups of differnt departments ?¶

Null hypothesis ($H_0$) : There is no significant difference among the groups.¶

alternative hypothesis ($H_a$) : There is a significant different among groups.¶

Recap:¶

Job Role/Level Features:¶

Which job role has the highest attrition rate and which has the lowest?¶

Which level role has the highest attrition rate and which has the lowest?¶

Is there a significant difference among groups of differnt Job roles/Levels ?¶

Null hypothesis ($H_0$) : There is no significant difference among the groups in considering numerical value.¶

alternative hypothesis ($H_a$) : There is a significant different among groups.¶

The test shows that there is a significant difference between groups from different job roles/levels in values such as:¶

For example the Above graph showes that the distribution Ages from some of job levels is significantly different which agree with ANOVA result¶

Recap:¶

Monthly Income:¶

To know if the Monthly income plays significant role in attrition¶

Null hypothesis ($H_0$) : Monthly income does not affect the attrition.¶

alternative hypothesis ($H_a$) : Monthly income has a significant effect on attrition.¶

Let's execute ANOVA test to know what categorical variables depend on monthly income¶

ANOVA test results showes that there is significant difference in monthly income between defferent groups of¶

Department, Education, Job levels, Job roles¶

The next graphs agree with what we conclude from ANOVA test:¶

Since $F > F_c$ or $P< 0.05$ so we can say that the mean of monthly income is defferent from department to another.¶

We can see significant differece between defferent departments and also between employees who quit and others who stayed in monthly income.¶

Recap:¶

Over Time :¶

Is overtime option dependent on specific departments?¶

Are employees who work overtime get paid more than others who don't?¶

Recap:¶

Features related with experience:¶

Does the number of previous companies worked for affect the attrition rate?¶

To know if the number of previous companies worked for affect the attrition¶

Null hypothesis ($H_0$) : Number of previous companies worked for does not affect the attrition.¶

alternative hypothesis ($H_a$) : Number of previous companies worked for affect the attrition.¶

To know if years of service in different acpects affect the attrition¶

Null hypothesis ($H_0$) : Attrition is independent of years of service.¶

alternative hypothesis ($H_a$) : Attrition is dependent on years of service.¶

Years since last promotion seems that has no effect on attrition but other feature related with working years significantily affect attrition¶

Recap:¶

Satisfaction rates:¶

What is the impact of Satisfaction rate on Employee attrition?¶

To know if level of satisfaction has a relation with monthly income.¶

Null hypothesis ($H_0$) : level of satisfaction does not depend on monthly income.¶

alternative hypothesis ($H_a$) : level of satisfaction has a relation with monthly income.¶

Recap:¶

Considering Personal reasons ¶

Age Analysis:¶

Is there a significant association between age and attrition?¶

Is there a significant difference between ages of males and females overall/between active employees and others who quit?¶

young employees leaves the company more compared to elder employees may me they seek for better experience or more monthly income.¶

Are elder get more monthly income?¶

The red area define the prediction Interval region PI.¶

Corrlation statistis $P_r= 0.5$ which indicate that there is a moderate positive correlation between Monthly income and age.¶

Recap:¶

Gender Analysis:¶

Does Gender has a significant effect on employee attrition?¶

Education:¶

From which eduactional background employees left the organization more?¶

Recap:¶

Marital Status:¶

Distance from home:¶

Work_life balance:¶

Recap:¶

Statistical Analysis ¶

Correlation matrix:¶

Recap:¶

We will use this analysis when training a model to drop some features.¶

We will formulate our hypothesis as:¶

Null hypothesis $H_0$ : There is no association between considering numerical feature and Employee Attrition¶

overview
¶

Considering Technical reasons
¶

`Null hypothesis` ($H_0$) : There is no significant difference among the groups.¶

`alternative hypothesis` ($H_a$) : There is a significant different among groups.¶

`Null hypothesis` ($H_0$) : There is no significant difference among the groups in considering numerical value.¶

`alternative hypothesis` ($H_a$) : There is a significant different among groups.¶

`Null hypothesis` ($H_0$) : Monthly income does not affect the attrition.¶

`alternative hypothesis` ($H_a$) : Monthly income has a significant effect on attrition.¶

Let's execute `ANOVA` test to know what categorical variables depend on monthly income¶

`Department`, `Education`, `Job levels`, `Job roles`¶

`Null hypothesis` ($H_0$) : Number of previous companies worked for does not affect the attrition.¶

`alternative hypothesis` ($H_a$) : Number of previous companies worked for affect the attrition.¶

`Null hypothesis` ($H_0$) : Attrition is independent of years of service.¶

`alternative hypothesis` ($H_a$) : Attrition is dependent on years of service.¶

`Null hypothesis` ($H_0$) : level of satisfaction does not depend on monthly income.¶

`alternative hypothesis` ($H_a$) : level of satisfaction has a relation with monthly income.¶

Considering Personal reasons
¶

The red area define the prediction Interval region `PI`.¶

Statistical Analysis
¶

`Null hypothesis` $H_0$ : There is no association between considering numerical feature and Employee Attrition¶

`alternative hypothesis` $H_a$ : There is a significant different and this numerical feature affect Attrition¶

`Null hypothesis` $H_0$ : There is no different in considering numerical feature between employees who left¶

`Alternative hypothesis` $H_a$ : There is a significant different between two groups¶

`Null hypothesis` $H_0$ : There is no association between considering categorical feature and Employee Attrition¶

`Alternative hypothesis` $H_a$ : There is a significant different and this categorical feature affect Attrition¶

predictive Model
¶

Bootstrap confidence interval
¶

Risk Level
¶

Features importances
¶

Interactive Dashboard
¶